On Jan 29, 2026, at 03:41, John Levine <[email protected]> wrote: > > While that is a tour de force of ASCII art (UTF-8 art?) it impresses > me as exactly what we should *not* do.
The objective needs to be that an RFC is accessible, not that each of the renderings we provide be accessible on its own. The current situation is untenable for accessibility because we render the math *before* creating the XML. That creates high fidelity (at least as long as you have the fonts, which we can deliver with the PDF, need to reference from copies controlled by us with HTML, and have no control over with TXT), but it loses the *source*. There is no actual representation of the math in an RFCXML file today! That needs to be fixed, and I applaud the effort to make the actual math available in the RFCXML. It will then be the HTML that provides the best choices for an accessibility environment to provide additional accessible renderings. (The same kind of choice can be applied to other pieces of rendered structured information; call me strange, but my brain would rather read UML source than look at Figure 2 in the TXT rendering of RFC 9880-to-be [0].) (Note that a TXT-rendered formulae can be quite accessible as long as the rendering is 1-D; it is the 2-D rendering that breaks access. I think we should nudge the tool providers to get better in doing 1-D renderings, as they are pretty much needed for inline math anyway, and in indicating whether there is a good 1-D rendering. Figure 5 of RFC 9438 [1] does rely on 2-D, as do several unnumbered pieces of block math, so 1-D is not a panacea.) The other kind of “accessibility” that is often cited as a goal is access for search tools. It would be good to be able to search for text snippets inside the math (which is possible today for TXT but not for HTML/PDF), but there is little hope to search for structure like the one in the unnumbered math block that follows Figure 2 of RFC 9438 [2]. Again, preserving source to the rescue. Grüße, Carsten [0]: https://www.rfc-editor.org/authors/rfc9880.txt#line=469 [1]: https://www.rfc-editor.org/rfc/rfc9438#figure-5 [2]: https://www.rfc-editor.org/rfc/rfc9438#figure-2 -- rswg mailing list -- [email protected] To unsubscribe send an email to [email protected]
